Resource agents described in documentation, available at https://raw.github.com/madkiss/openstack-resource-agents/master/ocf are somehow unnecessary since Pacemaker supports LSB/Upstart which should work on all up to date Linux distributions.
All three resource agents from above link are faulty and require certain amount modification, depending on distribution. Especially Neutron L3 Agent where author of resource agent presumed communication between L3 agent and Neutron server is direct which is not correct as communication occurs over RabbitMQ (or similar) service. This makes Neutron L3 agent completely unusable and needs to be heavily modified for it to be able to work.
Clean and simple Pacemaker configuration with upstart resource configured looks something like this:
primitive p_neutron-dhcp-agent upstart:neutron-dhcp-agent \
op monitor interval="30s" timeout="30s"
primitive p_neutron-l3-agent upstart:neutron-l3-agent \
op monitor interval="30s" timeout="30s" \
op start interval="0" trace_ra="1" \
op restart interval="0" trace_ra="1"
primitive p_neutron-metadata-agent upstart:neutron-metadata-agent \
op monitor interval="30s" timeout="30s"
group g_services_network p_neutron-l3-agent p_neutron-dhcp-agent p_neutron-metadata-agent \
meta target-role="Started"
With these in place there is no need for resource agents or anything similar.
-----------------------------------
Built: 2014-07-28T07:14:27 00:00
git SHA: 10106b24fc90111a0617f6e94d8a83c57520c6d6
URL: http://docs.openstack.org/high-availability-guide/content/ch-network.html
source File: file:/home/jenkins/workspace/openstack-manuals-tox-doc-publishdocs/doc/high-availability-guide/ch_network.xml
xml:id: ch-network
I believe LSB/upstart agents could be used only for demonstration purposes. There shoud be a Note about this in the guide. In reality, some additional high level checks might be required for resources and OCF fits the best for it. These checks could be for AMQP connections, API responsibility, application health (for example, if the user wants multistate clones for messaging or DB) and so on.
Anyway, putting this checks from OCF to upstart scripts shipped with the packages, would not simplify code supporting tasks. Some one (a guess, an Openstack and Linux HA community) should maintain then upstart scripts as well as OCF.